RESULTS

Summary of Sedimentary Environment

Middle Valley sediments are composed of hemipelagic silty clays with occasional turbidite sequences (Shipboard Scientific Party, 1998). The sediments recovered at Site 1036 show the influence of the nearby venting of 270°C hydrothermal fluid (Butterfield et al., 1994) in their high temperatures (Shipboard Scientific Party, 1998). However, in the intervals sampled for this study (Table T1), all three holes were experiencing hydrothermal recharge. Recharge was demonstrated by direct measurements of negative pore pressures in the upper layers of sediment and is also seen in the chemistry of the pore fluids (Shipboard Scientific Party, 1998). Pore fluids above 20, 40, and 50 mbsf at Holes 1036A, 1036B, and 1036C, respectively, are consistent with seawater that has been conductively heated and subsequently reacted with the surrounding sediment (Shipboard Scientific Party, 1998). Therefore, these sediment samples have markedly different chemistry than typical hydrothermal samples in the same temperature range (60°-130°C), which are thought to achieve their intermediate temperatures through mixing of seawater and hot, reduced hydrothermal fluids.

Direct Counts of Microorganisms

The number of microbes was determined from formalin-fixed samples for all whole-round cores where complementary lipid analyses were performed. Objects counted as cells under Acridine orange appeared to be small (0.2-0.4 µm), green-fluorescing coccoids; most but not all of these also stained with DAPI. Counts were extremely low in all samples (Table T1), and the detection limit for direct counts in this study was 1.9 x 10⁵ cells/cm³, which amounts to one Acridine orange-stained bacterially shaped object per 1500 microscope fields. Some samples had bacterial densities below this detection limit, but even in samples above the detection limit, fewer than 20 cells were observed per prepared slide. The 95% confidence intervals for these counts are estimated at ±0.3 log units.

Lipid Analyses

PLFAs were detected at low levels in all samples (Table T1), although Sample 169-1036C-4H-5, 140-150 cm, was near the detection limit of 0.5 pmol/g. Eubacterial abundances corresponding to these PLFA contents were calculated using a conversion factor of 2.5 x 10⁴ cells/pmol PLFA (Balkwill et al., 1988) and an average factor of 1.2 g (dry weight) of sediment per cm³ (Shipboard Scientific Party, 1998). All PLFAs (including very long-chain saturated fatty acids) were included in the biomass estimate. No ether-linked lipids were detected; the detection limit for ether-linked lipids is also 0.5 pmol/g.

Detected PLFAs are shown in Table T2 as mole percentages. Sample 169-1036C-4H-5, 140-150 cm, is excluded from this summary because the total PLFA in this sample was so near the detection limit that only the very most abundant PLFAs were observed. The normal saturates 16:0 and 18:0 were the most abundant PLFAs in these sediments, accounting for 44%-65% of all detected PLFA. Monounsaturates 16:17, 18:17, and 18:19 formed the second most predominant group (21%-34%) and branched-chain saturates (i15:0, a15:0, i16:0, i17:0, and a17:0) were 11%-23% of the total PLFA. Cyclopropyl fatty acids cy17:0 and cy19:0 were 3%-9% of the total PLFA. No polyunsaturated, branched-chain unsaturated, or trans-monounsaturated PLFAs were detected in these samples. Very long-chain saturated PLFAs (VLCFAs, greater than 20 carbons) were also measured in quantities comparable to those that contained 20 carbons or fewer (Table T1).

Samples were divided on the basis of temperature into warm (Samples 169-1036B-2H-1, 140-150 cm, and 169-1036C-2H-5, 130-140 cm), hot (Samples 169-1036B-2H-5, 140-150 cm, and 169-1036C-3H-5, 140-150 cm), and very hot samples (Samples 169-1036A-2H-1, 140-150 cm, and 169-1036C-4H-5, 140-150 cm). PLFA-derived biomass was not significantly different between warm and hot groups (t-test, p > 0.5), although VLCFA abundance did vary significantly between those two groups (t-test, p < 0.02). To test the hypothesis that temperature was the main factor controlling community composition, arcsine-transformed PLFA mole percentages (including VLCFAs) were subjected to K-means cluster analysis using SYSTAT (Wilkinson et al., 1992). The K-means analysis splits samples into a user-specified number of groups and swaps samples among the groups until between-group variation has been maximized (Hartigan, 1975, as modified by Wilkinson et al., 1992). K-means analysis on this set of samples failed to divide the samples into groups; instead single samples split off the main group as the number of groups was increased.

PLFA profiles (mole percentages) from Middle Valley sediments were compared with other PLFA analyses of samples from deep or hot environments (Fig. F1). Cluster analysis was used to gauge the similarity and relationship between samples; for this comparison, VLCFAs greater than 24 carbons in length were excluded from the analysis as they are not routinely measured in all samples. Euclidean distances and Pearson correlation coefficients (Pearson's R), joined with average value cluster linkage in SYSTAT (Wilkinson et al., 1992), gave similar branching topologies (Fig. F2). Hot samples (Samples 169-1036B-2H-5, 140-150 cm, and 169-1036C-3H-5, 140-150 cm) clustered apart from the rest of the Middle Valley sediments, and this clustering was mainly driven by differences in the amounts of VLCFA. If VLCFAs were excluded from the analysis, the partitioning between Middle Valley samples became much less robust (Fig. F2B). Middle Valley sediment samples form a cluster distinct from hot sediments from Guaymas Basin (Guezennec et al., 1996) and deep shales and sandstones from New Mexico (Ringelberg et al., 1997). One very hot sample of active sulfide (Hedrick et al., 1992) from the Endeavour Segment, Northeast Pacific, was similar to Middle Valley sediments, although other sulfide samples were less similar.

Discussion

Middle Valley sediments do not provide an especially enticing microbial habitat. These sediments are not rich in microbial energy sources, unlike many hydrothermally influenced areas (Baross and Deming, 1995; Jannasch, 1995). Potential electron donors in the sediments include organic carbon (0.2 wt%), methane at about 10 ppm, and ammonium (Table 1; Shipboard Scientific Party, 1998). Oxygen and nitrate, favored electron acceptors, were not measured; although, by comparison with other sediments, at 10 mbsf oxygen is certainly depleted. Sulfate is present at all depths in all samples at concentrations 50%-100% of that in seawater (Table T1). These sediments, although hydrothermal in proximity, are unlike other hydrothermal locales previously sampled for microbiology and, in many respects, are much more comparable to terrestrial deep environments.

A few basic predictions can be made about the composition of the microbial community based on the physical and chemical environmental parameters. These sediment samples are 60°-130°C, and, therefore, any active community should be composed of high-temperature microorganisms. Sulfate is present in all samples and, because sulfate reduction is a dominant anaerobic lifestyle in marine sediments (Jørgensen, 1982), sulfate-reducing microbes should be prevalent. Any heterotrophic organisms that are active are probably oligotrophic, adapted to growing under conditions of low nutrients.

Biomass Estimates

Bacterial densities were extremely low but measurable (in most cases) in sediments from all three holes at Site 1036 (Cragg et al, Chap. 2, this volume). All intervals sampled in this study were near the detection limit of Acridine orange direct counts (this study: 1.9 x 10⁵ cells/cm³) and some were near the detection limit of PLFA (0.5 pmol/g sediment). Biomass estimates from PLFA analyses are usually presumed to represent viable cells, as phospholipids have been demonstrated to have extremely short half-lives in near-surface sediments (White et al., 1979b), likely because of phospholipase activity (Harvey et al., 1986). However, it is possible that an environment with very low microbial activity might preserve PLFA in an atypical fashion, so these biomass estimates must represent an upper boundary. Free fatty acids have been synthesized abiotically under hydrothermal conditions in the laboratory (McCollom et al., 1999), but these should not interfere with the biomass estimates because the extraction procedure used in this study specifically separates free fatty acids from PLFA.

The direct counts and PLFA estimates of biomass agree to within half an order of magnitude, which is relatively close considering the uncertainty in the direct counts. Agreement between direct counts and PLFA biomass estimates is variable in other studies pertaining to the subsurface: in some studies they agree within an order of magnitude (Balkwill et al., 1988; Fredrickson et al., 1995), whereas in others there is an order of magnitude offset (Kieft et al., 1994), or even two (Haldeman et al., 1995). Cell densities calculated from PLFA are vulnerable to the choice of conversion factor (PLFA per cell). The conversion factor chosen, 2.5 x 10⁴ cells per pmol PLFA, was calculated for subsurface cells with similar diameters (Balkwill et al., 1988). Other studies have used conversions as large as 5 x 10⁵ cells per pmol PLFA (Fredrickson et al., 1995).

The concordance between PLFA biomass and direct counts in this study supports previous measures of microbial biomass in deep sediments using Acridine orange direct counts (Cragg et al., 1990, 1992; Cragg, 1994; Cragg and Parkes, 1994). Although bacteria observed in direct microscopic counts are not necessarily active, PLFA in samples imply intact lipid membranes and at least indicate some form of microbial survival. The biomass found in Middle Valley sediments is comparable to biomass found in other hot, energetically depauperate environments. Piceance Basin in Colorado shows similar PLFA biomass (10 pmol/g) at 45°C, though this biomass decreases with depth and temperature (Colwell et al., 1997). Other very high-temperature environments where PLFA biomass has been measured include samples of active sulfide structures (10-370 ng/g, temperatures estimated at 50°-350°C) (Hedrick et al., 1992) and near-surface sediments in zones of diffuse hydrothermal upflow (200 pmol/g, temperatures estimated at 60°C) (Guezennec et al., 1996); these samples have the benefit of hydrothermal energy sources and could be expected to have larger biomass than the Middle Valley samples.

Community Composition

Lipid analysis can give insight into the composition of the microbial community. Whereas most organisms cannot be individually identified from a lipid profile, some classes of microbes have distinctive lipids (e.g., the Kingdom Archaea), and the presence or absence of these biomarkers is an indicator of the presence or absence of these classes. Other types of distinctive lipids are associated with the state of the cell and can be used as stress or growth indicators.

Surprisingly, archaeal diether and tetraether lipids were below the detection limit in all samples. The strong acid methanolysis performed on the whole sediment samples is designed to remove ether lipid residues from a solid matrix (Hedrick et al., 1992), and it is unlikely that there were ether lipids in the samples that went undetected. As the detection limit for the PLFA analysis and the ether lipids are similar, this implies that the archaeal biomass was less than 10% of the bacterial biomass in all samples. Most known organisms that grow in the temperature range 80°-110°C fall in the Kingdom Archaea (Brock et al., 1994), and archaeal lipids made up the bulk of the total microbial lipids in samples of hot sulfide (Hedrick et al., 1992). However, in these Middle Valley sediment samples, archaea were not a significant fraction of the microbial community. Preliminary analyses of hot, hydrothermally influenced surface sediments from Guaymas basin also showed no detectable archaeal biomass (Guezennec et al., 1996), but it is possible that study missed ether lipids present in the sediment because they did not use the robust strong acid methanolysis ether lipid extraction protocol.

Specific PLFAs are characteristic of different classes of Eubacteria. Monounsaturated PLFAs like the 16:17, 18:17, and 18:19 found in this study are indicative of gram-negative bacteria, whereas the terminally branched saturates i15:0, a15:0, i16:0, i17:0, and a17:0 are usually found in gram-positive bacteria or anaerobic microbes (White et al., 1996). The normal saturates 16:0 and 18:0 are abundant in all microorganisms.

Cyclopropyl PLFAs are generally found in gram-negative bacteria, and their abundance tends to vary with growth phase. As gram-negative bacteria enter the stationary phase of growth, they increasingly convert the monounsaturates 16:17 and 18:17 into the cyclopropyl fatty acids cy17:0 and cy19:0 (Guckert et al., 1985; Kieft et al., 1994). Cyclopropyl fatty acids are important components of the phospholipids in these sediment samples, and the cy17:0/16:17 and cy19:0/18:17 ratios range from 0 to 0.45 and 0.27 to 0.71, respectively. These ratios are midway between ratios seen in actively doubling cultures (0.05 or less) and cultures in stationary phase (2.5 or more) (Guckert et al.,1985); this may indicate that the sedimentary microbes are still actively growing to some extent.

In these sediment samples, a number of specific PLFAs are notable by their absence. No polyunsaturated PLFAs were detected, which are found in microeukaryotes (White et al., 1996) and some barophilic bacteria (DeLong and Yayanos, 1986). No trans-monounsaturated PLFAs were detected; these tend to be produced from the corresponding cis isomer during periods of stress (Kieft et al., 1994; Findlay et al., 1990a), including starvation stress. However, a subsurface microbe (Arthrobacter sp.) did not accumulate trans-PLFA in starvation experiments (Kieft et al., 1994; Kieft et al., 1997), so the ratio of trans- to cis-PLFA may not be a robust measure of stress in all environments.

No branched monounsaturate or mid-chain branched saturate PLFAs were detected. Desulfovibrio and Desulfobacter, two of the most widely distributed and easily cultured sulfate-reducing genera, have distinctive terminally branched monounsaturates (i17:17) and mid-chain branched saturates (10Me16:0), respectively (Kohring et al., 1994). Desulfotomaculum spp. also show the branched monounsaturate i17:17 (Kohring et al., 1994; Liu et al., 1997). Typically, these PLFAs make up 10%-20% of the total in cultured representatives of these genera (Kohring et al., 1994; Liu et al., 1997); therefore, as these PLFAs were not detectable in the sediments sampled, these genera cannot be dominant members of the microbial community. Whereas the absence of these biomarkers does not mean that all sulfate-reducing bacteria are absent from these sediments, these three genera are believed to be abundant in deep environments. Thermophilic Desulfotomaculum spp. have been isolated from deep, hot boreholes in the Taylorsville Triassic basin (Liu et al., 1997). Both Desulfovibrio and Desulfobacter have been identified in oil wells (Magot et al., 1992; Brink et al., 1994). All cultures of sulfate-reducing bacteria isolated from deep marine sediments (Bale et al., 1997; Barnes et al., 1998) are Desulfovibrio spp. by 16S rRNA sequence analysis. Hot sediments may have different suites of sulfate-reducing microbes, though; sulfate reduction was measured in hot sediments from Guaymas Basin to temperatures of 110°C (Jørgensen et al., 1992), and pushcores taken at Guaymas do not show any i17:17 or 10Me16:0 (Fig. F1). Sulfate reduction at these temperatures might be expected to be moderated by Archaeoglobus or other sulfate-reducing archaea (Baross and Deming, 1995).

VLCFAs were detected in these sediment samples (21:0 through 30:0). Long-chain saturates of 20-23 carbons were found in lipids from sediments collected at the Guaymas basin hydrothermal area (Guezennec et al., 1996) and have been noted in cultured isolates of oligotrophic soil bacteria (Rezanka et al., 1991). VLCFAs are not necessarily routinely analyzed with other phospholipids (as the gas chromatography run may be terminated before they elute), so the unreported VLCFAs in other studies of PLFAs cannot always be taken as evidence of absence. These lipids are more abundant in the two 100°C samples than in the two 60°C samples, but they are a large fraction of the total lipids (16%-49%) in all five samples. The significance or function of these very long-chain saturated PLFAs is not known, but they could be a signature of oligotrophic lifestyles or perhaps involved in thermal protection of membranes, as increased chain length raises the melting point of lipids.

Community Comparisons and Dynamics

Whereas some classes of organisms have distinctive lipids and can be identified directly from a lipid profile, most organisms will only contribute to the overall pattern. This profile of lipids, or lipid fingerprint, will be different for different microbial communities, and comparison of lipid fingerprints over time or between locations to determine differences in microbial community structure is the most powerful way to use environmental PLFA data. PLFA profiles have been used to compare microbial community structures in a range of sediments and rocks: boreal peatlands (Sundh et al., 1997), wetland sediments (Boon et al., 1996), field manipulations of marine sediments (Findlay et al., 1990b), contaminated sediments of marine bays (Rajendran et al., 1994), deep sandstones and shales (Ringelberg et al., 1997), hot surficial sediments near deep-sea hydrothermal vents (Guezennec et al., 1996), and metal sulfide precipitates from deep-sea hydrothermal vents (Hedrick et al., 1992).

The microbial communities in the sampled sediments from Middle Valley are relatively similar to one another, regardless of temperature differences. K-means analysis showed that communities did not readily partition into distinct groups. Cluster analysis grouped the two hot samples (Samples 169-1036B-2H-5, 140-150 cm, and 169-1036C-3H-5, 140-150 cm) together, though this difference was mainly driven by the distribution of VLCFAs. Removal of VLCFAs from the data set increases the self-similarity of this group (Fig. F2B). VLCFA showed an increase from warm (60°C) to hot (100°C) samples, though the VLCFA in the hottest sample (130°C) was much lower than either the warm or the hot samples.

Comparison of Middle Valley PLFA community profiles with PLFA profiles from other deep or hot environments (Figs. F1, F2) shows that these deep marine sediments are more similar to each other than they are to other sampled environments, with the exception of an extremely hot sample (100°-350°C) of sulfide (Hedrick et al., 1992). Pushcore samples taken in hot hydrothermally influenced sediments at Guaymas Basin (Guezennec et al., 1996) are more similar to deep cores of sandstone and shale from New Mexico (Ringelberg et al., 1997) than they are to the Middle Valley sediments. Guaymas sediments are influenced by the upflow of hydrothermal fluids and thus are expected to differ from Middle Valley sediments, which are similarly hot but experiencing downflow of seawater. Unexpectedly, Guaymas sediments intermingled with deep sandstones and shales in the cluster analysis.

If buried microbial communities at Middle Valley remain active at high temperatures, the low-temperature community active at the time of deposition must be replaced by a high-temperature community. Thermophilic replacements may have been dormant in the sediment (as was shown in Pacific abyssal sediments by Dobbs and Selph, 1997) or potentially advected into the hot region. Middle Valley sediments are mainly silty clays and, because downflow through the sediments is 10-100 cm/yr (P. Schultheiss, pers. comm., 1997), micrometer-sized particles such as bacteria are likely to have difficulty penetrating down through sediments of comparable grain size. There is no evidence for lateral advection at any of the depths sampled (Shipboard Scientific Party, 1998). It is likely that the microbes making up the communities in Middle Valley sediments were present at the time of burial, and any succession events were composed of microbes (active or inactive) initially deposited at the sediment/water interface.

These microbial communities do not show large or coordinated PLFA shifts with temperature, with the exception of the increase in VLCFA with temperature. This relatively consistent PLFA fingerprint implies that the overall structure of the microbial community has not changed or else microorganisms have been cryptically replaced with other microbes that have the same fatty acids. To determine which of these possibilities might have occurred, DNA extraction was attempted on preserved splits of the sediment samples, but no measurable or amplifiable DNA could be recovered. If a community shift occurred as temperature increased, thermophiles with similar lipid compositions have replaced mesophilic organisms. It is difficult to reconcile this scenario with the lack of detectable archaea in hot samples; based on our knowledge of high-temperature microorganisms, archaea should have dominated the biomass of samples in the thermal range 80°-110°C. A community shift to archaea was seen in the active sulfide structure samples analyzed by Hedrick et al. (1992), where, as the temperature increased, the ratio of ether-linked (archaeal) to ester-linked (bacterial) lipids increased from 0.1 to more than 20. Thus, it is likely that the organisms present at higher temperatures in Middle Valley sediments may be remnant communities from lower temperatures and therefore unlikely to be active. However, the caveat remains that our knowledge of prokaryotic diversity is meager (Pace, 1997), and it is possible that hot Middle Valley sediments contain previously unknown high-temperature, active bacterial communities that are similar in lipid composition to moderate temperature Middle Valley sediment communities.